prodmon

prodmon is a cross-platform desktop productivity monitoring application that uses AI to analyze user activity via periodic screenshots. It detects anomalies and deviations from role-based productivity baselines using both cloud (Google Gemini) and local (Gemma3 via Ollama) models.

Features

• Automated Screenshot Capture: Takes desktop screenshots at regular intervals.
• OCR Extraction: Extracts text from screenshots using Tesseract OCR.
• AI-Powered Anomaly Detection: Analyzes extracted text for productivity anomalies using:
  • Google Gemini API (cloud)
  • Local Gemma3 model (via Ollama)
• Role-Based Rules: Supports customizable rules and baselines for different job roles (developer, accountant, office worker, etc.).
• Learning System: Detects unknown entities in screenshots and updates rules using LLMs.
• Modern Electron Frontend: Real-time dashboard for monitoring, anomaly alerts, and model selection.
• Cross-Platform: Works on Windows, macOS, and Linux.

Screenshots

example screenshot 1: prodmon flags YouTube as unproductive

Project Structure

prodmon/
  src/
    index.js         # Electron main process (app entry)
    index.html       # Frontend UI
    index.css        # Styles
    logic/
      requirements.txt
      src/
        main.py      # Python backend entry point
        api/
          gemini.py  # Google Gemini API integration
          gemma3.py  # Local Gemma3 model integration
        utils/
          ocr.py     # OCR extraction logic
          cli.py     # CLI argument parsing and debug
          learn.py   # Learning system for unknowns
      rules/
        baseline.json
        developer/baseline.json
        accountant/baseline.json
        office_worker/baseline.json

Installation

Prerequisites

• Node.js (for Electron)
• Python 3.9+
• Ollama (for local Gemma3 model, optional)
• Tesseract OCR installed and in your PATH

Setup

1. Clone the repository:

   git clone <repo-url>
   cd prodmon

2. Install Node dependencies:

   npm install

3. First run (auto-setup):

   npm start

   • On first launch, the app will create a Python virtual environment and install backend dependencies automatically.

4. Configure Google Gemini API (optional for cloud AI):

   • Create a .env file in src/logic/ with:

     GOOGLE_API_KEY=your_google_gemini_api_key
     

5. (Optional) Setup Ollama and Gemma3 for local AI:

   • Install Ollama and pull the Gemma3 model:

     ollama pull gemma3:4b

Usage

• Launch the app with npm start.
• The app will take screenshots every 30 seconds and analyze them.
• Switch between "Gemini" (cloud) and "Local Model" (Gemma3) in the sidebar.
• View detected anomalies and activity stats in real time.

Customization

• Rules & Baselines:
  Edit or add JSON files in src/logic/rules/ for different roles or organizations.
• Add New Roles:
  Create a new folder in rules/ (e.g., rules/manager/) and add a baseline.json.

Dependencies

Node.js

• electron
• screenshot-desktop
• python-bridge

Python (see src/logic/requirements.txt)

• pytesseract
• Pillow
• opencv-python
• scikit-learn
• numpy
• pandas
• google-generativeai
• python-dotenv
• urllib3

Development

• Frontend: Edit src/index.html, src/index.css, and src/index.js.
• Backend: Edit Python files in src/logic/src/.
• Rules: Edit JSON files in src/logic/rules/.

License

MIT